Reliable quantitative data on the extent and rates of soil erosion are needed to understand the global significance of soil-erosion induced carbon exchange and to underpin the development of science-based mitigation strategies, but large uncertainties remain. Existing estimates of agricultural soil and soil organic carbon (SOC) erosion are very divergent and span two orders of magnitude. The main objective of this study was to test the assumptions underlying existing assessments and to reduce the uncertainty associated with global estimates of agricultural soil and SOC erosion. We parameterized a simplified erosion model driven by coarse global databases using an empirical database that covers the conterminous USA. The good agreement between our model results and empirical estimates indicate that the approach presented here captures the essence of agricultural erosion at the scales of continents and that it may be used to predict the significance of erosion for the global carbon cycle and its impact on soil functions. We obtained a global soil erosion rate of 10.5 Mg ha-1 y-1 for cropland and 1.7 Mg ha-1 y-1 for pastures. This corresponds to SOC erosion rates of 193 kg C ha-1 y-1 for cropland and 40.4 kg C ha-1 y-1 for eroding pastures and results in a global flux of 20.5 (±10.3) Pg y-1 of soil and 403.5 (±201.8) Tg C y-1. Although it is difficult to accurately assess the uncertainty associated with our estimates of global agricultural erosion, mainly due to the lack of model testing in (sub-)tropical regions, our estimates are significantly lower than former assessments based on the extrapolation of plot experiments or global application of erosion models. Our approach has the potential to quantify the rate and spatial signature of the erosion-induced disturbance at continental and global scales: by linking our model with a global soil profile database, we estimated soil profile modifications induced by agriculture. This showed that erosion-induced changes in topsoil SOC content are significant at a global scale (an average SOC loss of 22% in 50 years) and agricultural soils should therefore be considered as dynamic systems that can change rapidly. Copyright © 2011 John Wiley & Sons, Ltd.